Rakuten Mobile, Inc.

Paul Harvey

Pierre Imai

Tareq Amin

Rehmat Ullah

Mihail Yanev

Peter Kilpatrick

Di Wu

Ivor Spence

Blesson Varghese

Cooperative training migration is performed by training, cooperatively with a computational device through a network, the neural network model, creating, during the iterations of training, a data checkpoint, the data checkpoint including the gradient values and the weight values of the server partition, the loss value, and an optimizer state, receiving, during the iterations of training, a migration notice, the migration notice including an identifier of a second edge server, and transferring, during the iterations of training, the data checkpoint to the second edge server.

Cooperative training migration

A controller for an autonomous network includes a processor configured to execute composable and replaceable modules in an interconnected manner to configure controller components. The controller components comprise a sensing component configured to collect sensor data about at least one controlled element under control of the controller, an analyzing component configured to process the collected sensor data to derive a current state of the at least one controlled element, a deciding component configured to, based on the derived current state, decide an action to be made with respect to the at least one controlled element, and an acting component configured to carry out the decided action with respect to the at least one controlled element, by causing a change in at least one of an operation of the at least one controlled element, or a configuration of the at least one controlled element.

Autonomous network, controller and method

A method includes receiving a packet. The method further includes determining whether the packet is part of a responsive connection. The method further includes determining whether a responsive buffer is full in response to a determination that the packet is part of the responsive connection. The method further includes applying a responsive probability to the packet in response to a determination that the responsive buffer is full. The method further includes determining whether to drop the packet based on the responsive probability. The method further includes accepting the packet for processing in response to a determination that the responsive buffer is not full or in response to a determination not to drop the packet.

Flow queueing method and system

Growing community of inventors

Tokyo, Japan

Paul Harvey